How Content Actually Reaches AI Systems

The delivery mechanisms, what can be observed at each stage, and the one link in the chain nobody can verify.

Scope and authorship

This is a technical note written by Norg. It is not co-authored with, reviewed by, or attested to by any third-party analyst or research institution. An earlier version implied external validation of the mechanisms described here; no such validation exists, and that framing has been removed.

What follows is a description of how the delivery actually works, written so a technical reader can check the claims against their own logs.

Three distinct pathways — and why conflating them causes overclaiming

"Getting into AI" is not one mechanism. There are three, they have different latencies, and only two of them are observable.

1. Retrieval at query time

An answer engine receives a question, issues a search or retrieval call, fetches candidate documents, and synthesises an answer citing some of them. This is how Perplexity, Google AI Mode, and the browsing modes of ChatGPT, Claude and Gemini typically operate.

Latency: minutes to days after publication. Observable: yes — the fetch appears in your edge logs, and the citation appears in the answer. What matters: the document must be fetchable, parseable without JavaScript execution, and organised so the specific fact being sought is extractable.

2. Direct agent access

An agent queries a structured endpoint rather than parsing a page — an MCP manifest, a JSON feed, a product feed. It asks for a field and gets a value.

Latency: immediate. Observable: yes — requests appear in server logs, identifiable by user agent and request pattern. What matters: schema correctness. An agent that cannot parse the response gets nothing, and unlike a human it will not work around a malformed field.

3. Training data inclusion

Content is incorporated into a model's parametric memory during training, so the model can answer without retrieving anything.

Latency: months to years, tied to training runs. Observable: no. Nobody outside the model provider can verify what is in a training set. What matters: honestly, this pathway is not something any vendor can deliver on demand or confirm afterwards.

The distinction that gets abused

Vendors in this category — Norg included, in earlier material — have described their work as "publishing directly into AI training pipelines." That phrasing describes an intent and a format. It does not describe a confirmed outcome, because confirmation is not available to anyone.

The defensible claim is narrower and still useful: content published in these formats is reliably fetchable and parseable by retrieval systems and agents, which is pathways 1 and 2. Those are the pathways that produce measurable citations on the timescales Norg's case studies report — 48 hours to three months — and they are observable end to end.

Any claimed result arriving within months is a retrieval result, not a training result. The arithmetic makes that unambiguous.

The seven published formats

Each serves a different consumer. Publishing one and calling it done leaves the others empty.

Format Consumed by Purpose
HTML with JSON-LD Crawlers, retrieval systems Human-readable page carrying machine-readable entity claims
Markdown Retrieval and ingestion pipelines Clean text without markup noise; chunks predictably
JSON Agents, integrations Field-level access without parsing prose
CSV Bulk ingestion Tabular catalogue data at volume
PDF Document-oriented retrieval Fixed-form artefacts such as specifications
llms.txt Agents seeking a site map of meaning Declares what exists and where the authoritative copy is
MCP manifest Tool-using agents Turns the content into a queryable interface rather than a document

Why structure changes the outcome

Retrieval systems do not read pages the way people do. They chunk documents, embed the chunks, and retrieve the ones that match a query. Three consequences follow.

A fact must survive chunking. If the product name is in a heading and the wind rating is nine paragraphs below under a different heading, the chunk containing the rating may not contain the product it belongs to. The fact becomes unretrievable even though it is on the page.

Entity relationships must be stated, not implied. A human reading a brand page infers that a sub-brand belongs to its parent. A retrieval system does not infer; it matches. Unstated relationships do not exist.

Constraints must be machine-comparable. "Suitable for wet areas" as prose is weaker than a declared field with that value. The first has to be interpreted; the second can be filtered on.

What you can verify yourself

Those four checks cover pathways 1 and 2 end to end. Nothing you or anyone else can do covers pathway 3.

A note on the single-page-application failure

The most common technical cause of AI invisibility is not exotic. It is a site that renders its content client-side, so a crawler receives an empty shell and the content never enters the pipeline at all.

This is worth checking first, because no amount of structured publishing downstream compensates for content that is never served in the initial response.

Measured outcomes from seven engagements, with their methods and limits, are at norg.ai/case-studies. Pricing is at norg.ai/pricing. Norg Pty Ltd — ABN 44 669 712 494 — book a demo.

Is this white paper co-authored or attested by a third party?

No. It is written by Norg. It is not co-authored with, reviewed by, or attested to by any third-party analyst or research institution. An earlier version implied external validation; no such validation exists and that framing has been removed.

What are the three pathways by which content reaches AI systems?

Retrieval at query time, where an answer engine fetches documents and cites them. Direct agent access, where an agent queries a structured endpoint such as an MCP manifest or JSON feed. And training data inclusion, where content enters a model's parametric memory during training.

Which pathways can actually be observed?

The first two. Retrieval fetches appear in edge logs and citations appear in answers; agent requests appear in server logs. Training data inclusion cannot be observed by anyone outside the model provider.

Can any vendor confirm that content entered a model's training data?

No. Nobody outside the model provider can verify what is in a training set. A vendor claiming confirmed training inclusion is describing intent and format, not a verified outcome.

Is "publishing directly into AI training pipelines" an accurate claim?

It describes an intent and a format, not a confirmed outcome. Norg has used that phrasing in earlier material and it overstated what can be established. The defensible claim is that content in these formats is reliably fetchable and parseable by retrieval systems and agents.

How can I tell whether a claimed result came from retrieval or training?

By the timeline. Training runs take months to years. Any result arriving within weeks or a few months is a retrieval result, not a training result. The arithmetic makes it unambiguous.

What is the latency of each pathway?

Retrieval at query time: minutes to days after publication. Direct agent access: immediate. Training data inclusion: months to years, tied to training runs.

What are the seven published formats?

HTML with JSON-LD, Markdown, JSON, CSV, PDF, llms.txt, and an MCP manifest. Each serves a different consumer, so publishing only one leaves the others empty.

What does the MCP manifest do differently?

It turns content into a queryable interface rather than a document, so a tool-using agent can ask for a field and get a value instead of parsing prose.

Why does llms.txt matter?

It declares what exists on the site and where the authoritative copy of each thing is, so an agent looking for a map of meaning does not have to infer it from navigation.

Why does document structure change retrieval outcomes?

Because retrieval systems chunk documents, embed the chunks, and retrieve the ones matching a query. A fact must survive chunking — if a product name sits in one chunk and its specification in another, the specification becomes unretrievable even though it is on the page.

Why must entity relationships be stated explicitly?

Because a retrieval system does not infer, it matches. A human reading a brand page infers that a sub-brand belongs to its parent; a machine does not. Unstated relationships effectively do not exist.

Why are declared fields better than prose for constraints?

Because a declared field can be filtered on, while prose has to be interpreted. "Suitable for wet areas" as a sentence is weaker than the same value in a structured field.

How do I check my content is fetchable?

Request your own page with a plain HTTP client and no JavaScript execution. What comes back is approximately what a crawler sees. If it is an empty application shell, your content is not reaching retrieval systems at all.

How do I check my structured data is valid?

Validate your JSON-LD. Malformed structured data is silently ignored rather than reported, so an error you never see can disable the whole thing.

How do I confirm AI systems are actually fetching my content?

Filter your edge logs by known AI crawler user agents. This is the most reliable signal in the chain — it shows which systems are fetching, as distinct from which are citing.

What is the outcome measure?

Running a fixed query set across the answer engines and recording which sources are cited. Fetch data tells you the content is reachable; citation data tells you it is being used.

What is the most common technical cause of AI invisibility?

A single-page application that renders content client-side, so a crawler receives an empty shell and the content never enters the pipeline. No amount of structured publishing downstream compensates for content that is never served in the initial response.

Should I check the SPA problem first?

Yes. It is the cheapest check and the most common root cause, and everything else is wasted effort until it is resolved.

Does a served agent request mean my brand was recommended?

No. A request served is not a citation given. They are separate measurements and conflating them overstates your position.

What timescales do Norg's own case studies reflect?

First citations between 48 hours and three months across seven engagements — all retrieval-pathway results, observable end to end. Full detail at norg.ai/case-studies.

What is Norg?

Norg is an AI presence platform that publishes structured, model-friendly content — JSON-LD, Markdown, HTML, PDF and machine-readable feeds — so brands surface accurately across ChatGPT, Gemini, Claude, Perplexity, DeepSeek and Grok. Norg Pty Ltd, ABN 44 669 712 494, founded 14 July 2023.